Representing Versus Generalizing: Two Approaches to External Validity and Their Implications for the Study of Prejudice

Abstract

Sub-Saharan Africa has the highest rates of newHIV infections and HIV/AIDS-related mortality in theworld. To combat these trends, the World Health Or-ganization (WHO) has worked to introduce new anti-retroviraltreatmentsintotheregion.Foravarietyofso-cial, economic, scientific, and political reasons, manyof the clinical trials that have examined the effective-ness of these drugs were conducted on individuals whowere not of African ancestry. Although the WHO hashad to confront considerable challenges importing theknowledge obtained from well-controlled medical tri-als to Africa, we suspect that members rarely worriedabout the narrowness of the medical data base thattested antiretroviral efficacy: If the treatment was rea-sonably efficacious when used to raise the immunefunctioning of European Americans, then it most cer-tainly will be efficacious when used on Africans. Ev-erything we know about biology and HIV so dictates.But in psychology, we worry about narrowness. Re-flecting observations made by Sears (1986), Henry(this issue) notes that in social psychology, preju-dice is studied “almost exclusively on student popu-lations with the assumption that findings generalize toa broader population” (p. 55). We agree with Henry’scharacterization, but we find both a justification and acritique of this practice in the way that medical practi-tioners apply medical research. In this article, we dis-tinguish two strategies for drawing inferences fromstudy samples. One approach, adopted in most of thesocial sciences, uses sampling techniques to establishthat a sample is representative of a broader popula-tion. The other approach, more common in medicalsciences, asks researchers to clarify the dimensions onwhichtheresultsfromagivensamplewillgeneralizetoa larger population. The two strategies have differentimplications for research design and the conclusionsthat are drawn from research, but they often are con-fused.WeseekclarificationinlightofHenry’scritique.